[Bug ld/16139] ld/mips: abort in _bfd_merged_section_offset

[raphael.droz+floss at gmail dot com]

https://sourceware.org/bugzilla/show_bug.cgi?id=16139



Raphaël Droz <raphael.droz+floss at gmail dot com> changed:



           What    |Removed                     |Added

----------------------------------------------------------------------------

                 CC|                            |raphael.droz+floss at gmai
l dot co

                   |                            |m



--- Comment #2 from Raphaël Droz <raphael.droz+floss at gmail dot com>
 ---

Created attachment 7310

  --> https://sourceware.org/bugzilla/attachment.cgi?id=7310&action=edit

gdb commented trace of a crash



I have this issue on Gentoo AMD 64 that I heavily commented in 

https://bugs.gentoo.org/show_bug.cgi?id=487910

and which seems very related to bug #13797 and bug #11584.



I was able to build a -gdb enabled libbfd which exhibits the problem.



Most of what is could say is explained in the attached gdb backtrace.

But I can't tell why this rel->r_info is wrong in some case but not always.



Let me know if I can provide anything more like .s files or output of my

debug'ged build.

I can provide also provide a tar.gz of this files.



Some samples of wrong values are:

0x4000002600000000   

0y100000000000000000000000010011000000000000000000000000000000000

instead of:

0x2600000004          0y10011000000000000000000000000000000100



or

0x8000000000000000   

0y1000000000000000000000000000000000000000000000000000000000000000

instead of:

0x1a00000004          0y1101000000000000000000000000000000100



My test-bed is a set of 2 .a and 2 .o files from a dovecot compilation

directory but this happens consistently with other program like... binutils

themselves and gcc (which I was never able to compile since the issue arise)





I noticed the issue recently (some months) but was able to reproduce it with

2.21.1-r1 and 2.23.1.

(compilations option seriously affects the chances of hitting the issue)

The chances of hitting a segfault increase consistently during the uptime w
hich

usually end in a system reboot.



-- 

You are receiving this mail because:

You are on the CC list for the bug.